Scalogram Based Heart Disease Classification using Hybrid CNN-Naive Bayes Classifier

Abstract

The proper monitoring of ECG will help to identify patients with cardiac problems. In the last two decades, many lives have been saved due to the automated prediction of heart diseases with the help of ECG signals. This article proposes a hybrid CNN-Naive Bayes classifier for classifying Normal Sinus Rhythm, Abnormal Arrhythmia, and Congestive Heart Failure from the MIT-BIH arrhythmia database. The one-dimensional ECG signals are converted to two-dimensional scalogram images using continuous wavelet transform. The scalogram images eliminate noise filtering and conventional feature extraction steps that may lead to loss of beats. The proposed architecture uses GoogLeNet to extract independent and discriminating features, which aids the Naive Bayes classifier to attain a high accuracy of 98.76%.